Enter the 6-digit code from your authenticator app.
Claude analyzes the selected symbol from the bot's full market context — technicals (scanner score, 15m/1h/4h/1d structure, Quantum Ribbon & RSI squeeze, liquidity/FVGs/VWAP, order book), flow & positioning (CVD order flow, OI & funding, liquidation clusters, weekly-range & daily-candle reads, Connors RSI), regime (Markov per asset + macro proxies, regime-flip flag, Fear & Greed) and macro/news (curated Telegram feeds, daily briefs, econ calendar, prediction markets). It proposes a direction, a confidence-based size, and structure-based SL/TP — entering at market or a limit (limit fills are re-validated). You confirm before anything opens. Demo-only — never trades live.
Grounded in the bot's market context — and can run web searches for current/external info (e.g. news). No trades. Follow-ups keep the thread; switching symbol resets it.
How "Auto SL" calculates the stop loss.
Risk profile - set the same one as "Risk Profile" in Vectis Signals:
Candle timeframe for ATR calculation:
Anti-Stop-Hunt SL Offset
Shifts the auto-calculated SL by 3 ticks away from round-number clusters (e.g. BTC $80000.00 → $79999.97 for a LONG). Retail-stop clusters at psychologically strong levels are systematically harvested by VWAP-style algos (Samir Varma, Titans #150). Applied only to auto-computed SLs in Alert / Auto-Trade / Suggest-SL paths — manually typed SLs in the Edit panel are never modified.
Pre-Trade Rationale
Opens the 'Instructions for the agent' box on the New-Trade tab expanded instead of collapsed. Since 26.460 the box is always THERE either way, and what you type in it is always submitted — Off only means it starts folded. Forces a brief intent statement before manual entries, a coach-style anti-tilt anchor recommended in Titans #150 ('Three sentences are enough. Without it: no trade.'). The note appears on the Trade-Card, stays in History after close, and is the premise the pre-fill re-check judges a resting limit against — name a horizon there and it will not cancel on shorter-timeframe or scanner grounds. Only affects manual entries — Alert and Auto-Trades are not asked.
Used for Auto SL and Auto Size calculations.
Scores Volume ratio, ATR ratio, and Efficiency Ratio to recommend a position size multiplier (0.5x - 1.5x). Mirrors Vectis Signals logic.
When a recommendation is available:
Per symbol, not one number for the whole bot: scales the position by how volatile that asset is right now against its own last 90 days (EWMA forecast, RiskMetrics lambda 0.94). Calmer than normal sizes up, stormier sizes down, centred so it tilts around your risk setting instead of drifting above it. Measured on 3,047 closed trades: 13.6% less spread in results and 10.9% less drawdown, with the average result unchanged. Live trades only.
Scale Out: TP1 20% · TP2 30% · TP3 30% · TP4 20% | Scale + Runner: TP1-3 same sizes, TP4 runs with TSL | Full Runner: no TPs, closed by TSL or manually
Method:
Activate after:
Tolerance (ATR multiple before exit):
Conservative (0.10×) — tight buffer, quick exits, more whipsaw on wicks. Standard (0.20×) — balanced. Loose (0.35×) — wide buffer, survives wicks, later exits. Adaptive — starts from Loose (0.35×) and scales by volatility (ATR ÷ 50-bar ATR avg): high vol tightens to ≈0.18×, low vol widens to ≈0.46×. Floor 0.21×.
Confirmation bars (consecutive closes beyond SL):
Immediate exit when a candle body closes on the wrong side of the VWMA(100) baseline on the Trailing SL timeframe. LONG: close < baseline → exit. SHORT: close > baseline → exit. Bypasses ribbon SL tolerance and soft-confirm (classic NNFX rule — catches trend-structure loss before the ribbon-based SL fires). Applies to new trades from now on.
These modes move the SL to protect profits as TPs are hit — in addition to whichever trailing SL method is active above.
Break-Even: after TP1 → SL moves to entry. Step-Up: after TP1 → entry; TP2 → TP1; TP3 → TP2 etc.
POC is computed from the last 50 candles using OHLCV data. Decay weighting reduces the influence of older candles.
When On: each candle is weighted by 0.97i (i=0 = newest). Older bars contribute less.
Automatically partial-close a position when RSI crosses overbought (LONG) or oversold (SHORT). Requires a minimum profit to trigger.
After TP1 the swing SL ratchets up the 1h Quantum-Ribbon (EMA60) instead of parking at break-even — securing more of a runner as the trend extends. Ratchet-only: the stop never loosens (the autonomous routine may tighten it further, never reset it to break-even). Applies live to already-open swing runners too. Default On.
Max daily loss. Bot auto-pauses when exceeded. Resets at midnight UTC.
Max drawdown from equity peak. Auto-pauses bot. Manual reset required.
Max simultaneous open positions per mode. Live and each demo mode count separately. Trades with SL at or beyond break-even don't count (capital no longer at risk).
Max notional size per trade. Larger trades will be rejected.
Max simultaneous LONG or SHORT positions per mode. Counted separately for Live and each demo mode; break-even-protected trades excluded.
Cooldown after stop loss before a new trade in the same market is allowed.
Max new trades per day, counted separately for Live and each demo mode. Prevents overtrading in each mode.
Loss threshold per scope. Behavior controlled by Loss Response Mode below (0 = off).
After N consecutive losses (Cool-Down) or N losses per UTC day (Halt) the scope blocks new entries. Cool-Down inspired by trader-podcast 9th commandment "after 3 losing trades in a row, take a break" — circuit-breaker for mental reset instead of full-day ejection. Auto-resumes after duration + resets consec counter.
Cool-Down length in minutes (15-240, default 60). Active only when Loss Response Mode = Cool-Down. Counter resets to 0 when timer expires.
Max time allowed between two consecutive losing closes for them to count as the same streak. If more time passes, the streak restarts — so only losses clustered in a real adverse regime arm the Cool-Down, not unlucky trades scattered over many hours. Defaults scale per mode by trade frequency. 0 = off (streak never expires). The streak also resets at UTC midnight.
Before opening ANY demo trade (every demo mode) the live orderbook is walked at the intended position size; the trade is skipped if estimated entry slippage exceeds this. Size-aware — a thin book only blocks large orders, not small ones. 0 = off. This replaces the coarse QC 24h-volume floor as the primary liquidity guard.
Your threshold for MANUAL live market entries, in percent (the unit the Lighter UI uses). Above it the submit asks for confirmation and shows the measured figure — it never blocks you, and the estimate appears under the Size field while you type. Separate from the demo gate on purpose: that one guards UNATTENDED trades and stays strict. 0 = no notice.
After N consecutive wins in a scope, warn or auto-downsize. OkalaNQ: "When you're on a winning streak and you keep going, you're giving the market the opportunity to take it back." Symmetric to the Cool-Down on the left.
Number of consecutive wins to trigger. Counter resets on any loss. Default 5.
Warn = Telegram only. Downsize = Telegram + auto-reduce risk-per-trade by the % below for the duration on the right.
% of normal risk during downsize window (e.g. 50 = half risk). Active only when Response = Downsize.
How long downsize stays active. Auto-resumes after this, Telegram-confirmed.
Master toggle for repeat-alert suppression. When ON, two filters drop redundant same-direction signals before they reach the Alerts tab: (a) any open position in any scope (Live / Demo Signals / Demo Scanner) blocks further same-direction alerts on that ticker; (b) the cooldown on the right blocks same-direction repeats even without an open position. An opposite-direction alert lifts both locks (trend reversed → continuation signals are again worth seeing).
Candles of the alert's own timeframe after a same-direction alert before another same-direction alert of the same stream (source, ticker and timeframe) is accepted again (0 = disabled). Only active when 'First Signal Per Direction' is ON. 16 bars = 4 h on 15m and 16 min on 1m. An opposite-direction alert of the same stream resets the lock.
Blocks new auto-trades around high-impact macro events (NFP, FOMC, CPI, etc.). Data pulled automatically from the ForexFactory weekly feed — no manual entry required. Manual entries get a confirm-dialog instead of hard block.
Master toggle. When Off, no events block trades regardless of other settings.
Minutes before event start to begin blocking. Default 60 covers position-squaring choppiness.
Minutes after event to continue blocking. Default 30 covers initial news-spike volatility.
Which events trigger a blackout. High = NFP/FOMC/CPI level only (recommended).
Which currencies' events count. USD covers all major US macro releases.
Show a banner under the top-nav during an active High-impact blackout window. Default On. Doesn't change the gate — only a UI cue with quick Close-all / Cancel-limits buttons.
When BTC (crypto proxy) or SPY (equity proxy) moves sharply, instantly closes only the hurt side of correlated positions — a crash closes LONGs (shorts profit → kept), an up-spike closes SHORTS — before the Lighter cascade fills your stops worse. Then a re-entry cooldown + auto-resume. Applies to live + all demo modes; forex/commodity use a per-asset self-crash trigger. Master toggle Off by default.
Master toggle. Off = no triggers, fully inert.
Re-entry lock after a trigger, then auto-resume (covers the post-crash bounce).
Magnitude in the window (either direction). MSTR rides this proxy.
For stock-perps / ETFs. Polled from Yahoo during US market hours.
Per-asset for forex / commodity (no shared proxy).
Blocks US-equity auto-trades around Options Expiration windows. Monthly OPEX = 3rd Friday of each month. Quarterly OPEX (Mar / Jun / Sep / Dec) = Triple Witching with index futures + index options + stock options all expiring together — hedge volumes 3-5× larger, larger window. Market makers must unwind gamma hedges in the surrounding hours, causing atypical intraday volatility and pin-risk at popular strikes. Only equity / index symbols are gated — crypto, forex, commodity remain active.
Master toggle. When Off, OPEX windows are ignored.
Minutes before 16:00 ET close to begin blocking on Monthly OPEX. Default 30.
Minutes after 16:00 ET close to continue blocking. Default 60.
Window-size multiplier for Triple Witching (Mar/Jun/Sep/Dec). 2.0 → 60min pre / 120min post on those days.
When On, block the entire OPEX trading day instead of using pre/post window. Most users leave Off (Speaker warned about the hours around expiry, not the whole day).
Blocks auto-trades on equity, index, ETF, and commodity symbols when the NYSE is closed for a holiday (10 full days/year) and optionally during the 13:00-16:00 ET early-close windows (3 days/year). Crypto and forex are unaffected. Manual trades pass through unchanged. Deterministic calendar — no external feed; observed-on-weekday rule applied automatically (Sat-holiday → Friday before, Sun-holiday → Monday after).
Master toggle. When Off, NYSE closures are ignored.
Top-of-page banner during holiday + advance window. Block still applies regardless.
Hours before a holiday's block-start when the banner appears with countdown. Default 4.
Include 13:00 ET early-close days (3rd July, Black Friday, Christmas Eve). When On, block starts 18:00 UTC.
Blocks all automatic trades in every mode (every automatic path: Signals / Scanner / Live AUTO / Swing / Narrative / Stage + future) during the volatile window around the NYSE open. Default 08:00–10:00 ET — thin pre-market plus the fakeout-prone first ~30 min after the 09:30 open. Weekday-only, asset-agnostic. Manual trades pass through unchanged.
Master toggle. Default On.
Window start, US Eastern (DST-aware).
Window end, US Eastern (DST-aware).
⚠ End time must be after Start time — the blackout window is currently inactive.
During the market weekend (Fri 22:00 – Sun 22:00 UTC) auto-trades are limited: Forex/index/stock-perp oracles are stale and crypto orderbooks are thinner. Manual trades stay possible but a confirm-dialog warns on non-whitelisted assets.
Master toggle. When Off, auto-trades run with the same rules 7 days a week.
Hard cap on risk-per-trade for auto-trades during the weekend window. Affects Live AUTO + the Demo AUTO paths (Signals / Scanner; the archived IVB and Scalp modes no longer trade). 0 = no cap.
BTC, ETH, SOL, BNB, XRP, ADA, AVAX — deep weekend orderbooks.
DOT, LINK, NEAR, ICP — thinner but still acceptable weekend liquidity.
Force 100% close at R-multiple during the weekend window. Overrides the active TP-Model with a single target — and forces TSL on from entry (Quantum Ribbon, Loose tolerance) to protect mid-trade profits when the target isn't reached (range/reversal). Closed config: TSL Method/Tolerance/Activation/Bars are fixed in Cap mode, not user-tunable. Applies to Live + Demo Signals + Scanner Demo. (Demo IVB, which it never applied to, is archived.)
Target distance in R-multiples of SL distance. 0.5R = half reward (symmetric to the halved Risk Cap). 1.0R = classic 1:1 R:R. 1.5R = wider cap, more reward on longer moves.
Master toggle. When Off, Max Heat is not enforced.
Max sum of all risk-at-stop (entry to SL × size) relative to equity. Break-even-protected trades contribute 0. Counted separately per mode.
Master toggle for correlated-position checks (e.g. BTC + ETH). When Off, both modes on the right are ignored.
Behavior on user-initiated entries when a correlated position is already open.
Behavior on webhook-driven auto-trades when a correlated position is already open.
Measured book correlation (20-min snapshot, blended 3/6/12d daily returns). Fires on new entries when the book's mean or strongest pair reaches 0.70. Swing, Live and Structure books; Structure stays warn-only during its frozen-rule eval.
Paste into TradingView alert Webhook URL. HTTPS port 443. Only accepts requests from TradingView's IP addresses.
Automatically trade when a TradingView alert arrives for a configured asset.
Assets set to On will be traded automatically when a TradingView alert arrives. Uses Strategy tab SL/TP settings.
Assets on this list always bypass the liquidity filter — they will never be blocked due to transient low orderbook depth. BTC, ETH, and SOL are additionally hardcoded as always-liquid.
Paper trading with simulated fills. Use the account toggle in the navbar to switch between demo accounts. Each demo account tracks its own balance, trades, and history independently.
These settings apply to Demo Signals, Scanner Demo, and Demo IVB (archived) trades. When set to 'Use Strategy Settings', they use your Strategy tab defaults. Override to customize SL, TP, and trailing for demo testing independently from live. Demo Scalp, Demo QC, Demo QC AI, and Demo Structure (all archived since v213.801) are not affected — each has its own dedicated settings (Scalp uses the Mean-Reversion-Scalper config; QC runs its own 3m scalp SL/TP; QC AI uses the AI platform's payload SL/TP; Structure runs its frozen eval rule: payload SL, fixed 2R TP, 24h time-stop).
SL Calculation Method
Candle timeframe for ATR calculation:
⚠️ 3m best for Crypto + USDT-quote pairs (BTC/ETH/EURUSDT/XAUUSDT) where Binance/Bybit data is available. Lighter-only assets (USDCHF, USDJPY, WTI, NATGAS, US stock-perps) fall back to oracle-fed Lighter candles — Min-SL floor compensates but POC/ATR quality drops.
Anti-Stop-Hunt SL Offset
Risk per Trade
TP Model
Trailing SL
Method:
Activate after:
Tolerance (ATR multiple before exit):
Confirmation bars (consecutive closes beyond SL):
Auto SL Adjustment
NNFX Baseline Safety Net
Volume Profile POC Decay
Per demo mode, force the Quantum trailing-stop to arm at entry instead of after TP1 — independent of the global Demo Strategy Override above (and of any mode's own bracket, e.g. the Signals mode's NAS bracket which otherwise trails from TP1). On = tighter stops earlier → smaller losers, but more early stop-outs in the entry chop. Default Off everywhere.
Receives trades from TradingView webhook alerts (NAS Ultimate Algo). View trades by switching to Signals in the navbar.
Paste into a TradingView alert. Trades from this webhook land in the Signals Demo account.
Automatically execute demo trades when a TradingView alert arrives at the Demo Webhook. Writes to Signals Demo account.
When Off, alerts still appear in the Alerts tab but no demo trade is placed. Uses the Demo Webhook URL above and the same Symbol List from Shared Settings below.
HTF Candle Filter
Only takes LONGs while the higher-timeframe candle is green and SHORTs while it is red (the NAS Ultimate Algo v3 'Candles Overlay' idea). Trading with the daily trend filters out a band of lower-win-rate counter-trend entries; on lower HTFs (1h/4h) the filter tends to cut profit because signals often fire on reversals — recommended timeframe: 1d. Fails open when HTF data is unavailable. Auto-trades only; manual trades are never blocked.
NAS Bracket Mode
Replicates the NAS Ultimate Algo v3 bracket. The SL is placed at the recent swing low (LONG) / swing high (SHORT) — the structure stop the indicator uses — with a 4×ATR(RMA) fallback when candle data is unavailable. TPs at 0.5R / 1.0R / 1.5R (NAS's U/2U/3U ladder), ~80% booked across TP1-3 with a ~20% runner. The baseline safety-net is off for this mode (it tended to close fresh trades in the entry chop). At TP1 both Quantum trailing and a break-even SL step activate — a trade that reaches TP1 and then reverses scratches at entry instead of handing the booked TP1 profit back at the wider stop; the structure stop owns the phase up to TP1, then the ribbon trails the runner above break-even. When Off, the shared Strategy/Override settings apply.
Receives auto-trades from the scanner engine. Configure scanner auto-trade in Automation → Auto-Trade (Scanner Demo). View trades by switching to Scanner in the navbar.
Automatically execute demo trades based on scanner signals. Writes to Scanner Demo account only.
🔄 Runs server-side — works 24/7 without an open browser.
Trigger Mode
Position Size Mode
Vectis IVB Breakout indicator webhook signals (Fabio Valentini methodology). Indicator-driven exits — bot opens position with safety SL, then follows TP1/TP2/Re-Test/VWAP/Counter-Agg/Session-End exit webhooks from the indicator.
Master toggle for the IVB webhook receiver + trade execution.
Trend signals with confluence score below this are dropped at the receiver. Bounce/Reversal signals are unaffected (their payload score is always 0).
Block IVB open-signals while a position is already open on the same ticker+direction (any scope).
Drop same-dir IVB opens within this window. Default 0 (off) since the indicator has its own 10-bar internal cooldown.
Fallback safety-SL = N × ATR(14) on indicator's tf, used only if payload sl_distance and sl_level both invalid/missing.
Session Window — block new entries after this NYSE-time. Avoids Power Hour reversals + end-of-day position squaring with too little continuation runway.
Range 12:00–15:30 ET (NYSE close 16:00 ET). Default 15:00 = block the Power Hour. Exits (partial_close / close) are never blocked. Weekends auto-skipped.
Safety Nets — protection against missing exit webhooks from the indicator (e.g. TV down, Pine bug, network outage).
When Mark crosses +1.0R (TP1 level), auto-move SL to entry. Worst-case becomes break-even even if no exit webhooks arrive. Default On.
Force-close if trade open longer than this. 0 = disabled. Default 8h (~1 NYSE session). Unconditional age-based cutoff — designed for missing Pine close-webhooks but fires regardless.
Telegram alert when no IVB webhook for X minutes while position open. 0 = disabled. Default 30min. One-shot — re-armed by next webhook.
Force-close any IVB still open this many minutes past NYSE close (16:00 ET). 0 = disabled. Default 30min = 16:30 ET. Catches missing Pine session_end webhooks.
Configure in TradingView as "Any alert() function call" with this URL on each Vectis IVB Breakout chart you want to forward.
Full (1.0× size) = top-liquid (deep weekend orderbooks). Reduced (0.5× size) = thinner liquidity → half position size for lower dollar risk. Blocked = unsuitable for Fabio's method (forex without central volume aggregator, stable-pair too-flat range).
Vectis Mean Reversion Scalper indicator (1m mean-reversion BB-touch + RSI extreme). Pine drives full lifecycle: open → tp1_hit (60% partial + SL→BE) → tp2_hit (close runner) / be_hit / sl_hit / mr_complete.
Master toggle for /webhook/scalp receiver + trade execution.
Reject open-signals below this Pine strength (0–10). Indicator-side dashboard shows current per-bar strength.
Max parallel scalp positions. User decision 2026-05-21: 6 (fits Tier-S whitelist + Demo Signals + IVB + Scanner concurrency).
Scalp-specific risk-per-trade (separate from the Strategy-tab default). 1m stops are tight (0.05-0.10 %) → 3 % risk overshoots Lighter's per-market leverage cap. Default 0.5 % sizes XAU/ETH/BTC to ~$5 risk on a $1,000 account, well within 20× margin.
Fallback Stop % per asset class — used only when Pine omits sl_price (normally Pine supplies an explicit swing-based stop, so these are a safety net).
BTC / ETH / SOL
HYPE / mid-cap alts
NVDA / TSLA / SPY (Tier-A unlock)
XAU / XAG / WTI
Trade Management
% of position closed on Pine tp1_hit. Remainder runs at BE-SL. 1.0 = legacy full-close mode.
Bot-side reference only — Pine owns the actual TP2 trigger (swing-based level in the open-webhook).
Block opens when live (best_ask−best_bid)/mid exceeds this. Scalp can't survive wide spreads.
Force-close after this many minutes (safety net). Default 120min — sized for 5m TF MR setups (6-24 bars to resolve).
Telegram alert when no scalp webhook for X minutes while position open. 0 = disabled. Default 30min. One-shot — re-armed by next webhook.
Configure in TradingView as "Any alert() function call" on each Vectis Mean Reversion Scalper chart you want to forward (Tier-S whitelist: BTC / ETH / SOL / HYPE / XAU).
Initial Phase-1 Tier-S only. Tier-A unlock after 1–2 weeks of demo data. Anything not listed = blocked at the receiver.
QC Signals indicator (closed-source, quantcrawler.com). Plain-text BUY/SELL alerts via "Any alert() function call" → bot opens demo trades with adaptive ATR/POC SL/TP on the QC account (3m default). EXIT/Flip/Break events are separate alertcondition() alerts — not yet wired (Phase 2c). First-signal-per-direction + repeat-cooldown apply (QC fires repeatedly in a trend).
Master toggle for QC auto-trade. When Off, /webhook/qc still captures + shows alert cards but places no trade.
Per-Signal (default — like the indicator author / the TV strategy): every signal is its own independent trade, repeats open new trades (capped by max-concurrent). Required for the TV-strategy backtest to be comparable. Pyramiding: a repeat same-direction signal enlarges the open position (weighted-avg entry, while in profit + trailing active). The Add fields below apply only to Pyramiding.
Risk-per-trade for each add, sized to the current trailing-SL. 0 = use the Strategy risk-per-trade default.
Cap on add-tranches per position. QC fires often in a trend — prevents pyramiding into the reversal.
Skip QC trades on assets below this 24h quote-volume (Lighter). 0 = off. Default $30M ≈ XAU level.
Widens the tight 3m ATR SL (1.0× → 1.4× / 1.75×) on high-volatility assets so normal swings don't trip the stop.
ℹ️ Liquidity Floor filters out thin markets (slippage + noise stop-outs). Vol-Adaptive SL widens the tight 3m ATR SL on highly volatile assets (e.g. ZEC) so normal swings don't trigger the stop immediately. Both apply to QC only; the SL multiplier scales by the asset's ATR-as-%-of-price on the 3m bar.
TPs as R:R ratios off the SL distance, like the QC Signals indicator (author default: TP1 1R / TP2 2R, 2 TPs). 1 TP = 100% close at TP1. 2 TPs = the TP1 Close % closes at TP1, the rest rides to TP2. (Per-asset overrides come next; for now this applies to every QC asset.)
Comma-separated base symbols. When the toggle is On, QC only trades these; alerts on other symbols are still captured + shown but not traded. Off = trade all.
ℹ️ Keep this curated. Once the QC TV strategy is released you can backtest watchlist (and other) assets and add only the ones with a proven edge.
Configure in TradingView as "Any alert() function call" on each QC Signals chart (3m). For EXIT events (Phase 2c) add separate alerts via the named "QC Signals EXIT (…)" conditions to the same URL.
Auto-trades Vectis Structure open-candidates (ib_breakout + va_exit, 15m) from the existing /webhook/structure capture — under the frozen rule from the 2026-07-06 edge eval (+0.38R/trade backtested over n=345): only bias-aligned signals (indicator's own bias agrees with the side), crypto only (no XAU), va_loc = inside. Entry MARKET, SL = payload level (as % of the actual fill, slippage-robust, no min-SL clamp), single TP = fixed 2R, 24h time-stop. First-signal-per-direction; risk-based sizing. Capture keeps logging everything regardless of this toggle.
Master toggle. When Off, /webhook/structure still captures every alert but places no trade.
Autonomous Weinstein stage-2 + Ted-Zhang scope (US equity perps with deep history + BTC/ETH/SOL, gated to Lighter-liquid names), long-only — the short side failed its pre-registered OOS backtest. Daily entries off the stage loop: buy a fresh stage-2 uptrend (close > 21-EMA > 50-SMA > 200-SMA, fan-filtered), trim 1/3 at each extension threshold, exit on a daily close below the 21-EMA. GARCH/EWMA vol-targeted sizing, max 20 concurrent positions. Runs in DRY-RUN — logs its intended actions but opens/trims/closes nothing until go-live is armed here.
Dry-Run (default): the daily pass logs its intended buys / trims / exits but places nothing. Live: it opens real demo-scope Stage trades autonomously — still paper, no live-order path. To retire the mode entirely, kill it in the demo-mode archive above.
Long-term narrative/thesis scope (equity / ETF / commodity / crypto perps, weeks-months holds) — design frozen in docs/ws1_gate2_design.md. Entries will come from a biweekly LLM deep-research routine over a persistent thesis store: Layer A = Lighter demo perps with a mandatory invalidation stop, Layer B = TradFi/options recommendations (reported only, never bot-executed). Exit machinery: break-even at +1R, a single 25% de-risk at +2R, then a wide higher-timeframe trail; thesis invalidation is the primary exit. NOT the memecoin-narratives surface. Research/maintenance runners are not built yet — this section currently manages the paper account only.
Double gate: Enable turns the lane on (biweekly Opus research goes live: web + thesis-store writes; maintenance runs read-only). Trading only happens when Mode = Live (maintenance may open/close paper positions). Keep both OFF/Dry until the 19.07 arming + safety re-check.
Events: beyond the fixed weekly/biweekly slot, this reacts to three between-slot conditions — a thesis reaching its review date → a research re-evaluation; a held position bleeding funding, or its invalidation stop getting hit → a maintenance re-check (re-justify / confirm the thesis is broken). Off = only the fixed slot runs; the lane won't react between slots. Requires Enable to be On.
Research = biweekly (even ISO week), maintenance = weekly, both in the day+hour slot above. On a failed cadence run the bot self-retries up to 3× (45 min apart, headless — no metered API); if all fail you get a Telegram alert.
QuantCrawler AI automation (quantcrawler.com) — a SEPARATE mode from QC Signals above. The AI scans assets every ~13-16 min and POSTs a JSON setup on each entry. MARKET orders only. The bot uses the payload's SL/TP when the webhook sends them correctly (confidence>0 + plausible distances); otherwise it falls back to the configured percentages below (QuantCrawler's webhook is currently unreliable). Both are applied as % of the actual fill (slippage-robust). Single-TP full-close, risk-based sizing. First-signal-per-direction + repeat-cooldown apply.
Master toggle for QC AI auto-trade. When Off, /webhook/qcai still captures + shows alert cards but places no trade.
Skip setups whose payload confidence is below this (0-100 scale). 0 = off (trade every setup). A 0-1 fractional confidence is auto-scaled.
Min. minutes between same-direction entries on a ticker. 0 = off (fully QC-driven — recommended for scalping). First-signal-per-direction still blocks a 2nd concurrent same-dir position.
Fallback SL % (of fill), used only when the payload SL is missing/invalid (confidence 0 or placeholder). Match your QuantCrawler ticker config (Scalping default 0.1%).
Fallback TP % (of fill), single-TP full close, used only when the payload TP is missing/invalid. Match your QuantCrawler ticker config (Scalping default 0.2%).
Intraday Quantum-Ribbon trailing SL (resolution →). Single-TP has no TP1 anchor → it arms profit-based (threshold →). Until armed the fixed SL holds; then the SL ratchets behind the ribbon and locks gains if the trend reverses before the TP. Default Off — observe a few trades first (a tight trail can also cut winners that would have reached the TP).
Trailing starts once the trade is this many R in profit (R = entry→SL distance). 1.0 = after a full-R move (≈ halfway to a 2R TP). Higher = more room before trailing engages.
Candle resolution for the trailing ribbon and the RSI-exhaustion check (shared). Lower = reacts faster but whips on intraday noise; higher = smoother trend-trail but slower to catch a reversal. Default 15m.
Paste this as the webhook URL in QuantCrawler's automation alert. Expected JSON payload: ticker, action, direction, price, stop_loss, take_profit, quantity, confidence, reasoning.
Two ingestion paths into the same Demo EW account. (1) Manual paste — forward a Discord Elliott-Wave signal into the Telegram bot chat; the bot parses it (direction, entry-zone, absolute SL/TP1/TP2), shows an inline confirmation, and on your Yes opens the trade. (2) Webhook (URL below) — an external source POSTs the same signal text and it auto-trades (no confirm) using the default limit edge. Either way: price in the entry-zone → MARKET; price on the expensive side → PENDING LIMIT at the chosen zone edge (cancels on stop-invalidation / TP1-missed / 48h). Scale-out 50% TP1 / 50% TP2 + break-even stop after TP1, risk-based sizing. Demo-only.
Master toggle. When Off, pasted signals are still parsed + previewed but never traded.
Default for a PENDING limit when price is outside the zone. You can still override per-signal (near/mid/far) in the Telegram confirmation.
When On, a pasted message the structured parser can't read is sent to Claude to extract a self-contained entry (demo-only, still needs your confirmation; management / commentary / conditional triggers are ignored). Requires ANTHROPIC_API_KEY on the server + Elliott Wave Mode On. Cost is per-paste only.
POST the raw EW signal text (or a JSON body with a text field) here. Auto-trades when Elliott Wave Mode is On and the bot isn't paused — using the default limit edge above; no confirmation. The manual Telegram-paste path keeps working alongside it. The LLM Extractor toggle also applies to free-form webhook payloads.
Three independent paper accounts, each measuring one source against the Dexscreener price — no wallet, no key, no real trade, pure measurement. View under Degen in the navbar (the nav badge is the combined equity; the trades tab shows the per-lane balance strip).
Karoshi-AI Shadow-Scanner calls relayed to the Degen Webhook below. Exit ladder below applies to this lane.
When On, every new captured call opens a paper position (forward-only — never back-opens old calls). Off = inert.
Exit model (tunable)
TP1 sells 50% / TP2 25% / moonbag 25% trails the % off peak (armed after TP2); break-even stop after TP1; hard-SL is best-effort — fills at ~8% impact when the book is ≥ $5K at the fill, 30% on thinner/unknown books (v213.365); rug = total loss. Trailing % = how far off the peak the moonbag sells (50 recommended). Size % equity = each AUTO trade is sized at this % of the current SOL equity (start + realized), min 0.05 SOL.
Trading costs — GMGN (realistic)
Models the real GMGN round-trip cost so the paper PnL is honest. Platform % = GMGN's fixed 1% handling fee, charged on every buy and every sell (entry + each TP/SL fill). Priority SOL/tx = the priority + Jito anti-MEV tip paid per transaction (flat SOL, set dynamically on GMGN, ~0.002–0.006). Because it's a fixed cost it dominates at the small Size SOL of memecoin scalps — 0.003 SOL on a 0.1 SOL stake is 3% per tx, and a full laddered exit (entry + TP1 + TP2 + moonbag = 4 tx) burns ~12% before slippage. Entry price-impact is modelled separately from the pool's liquidity. (Lighter's 0% fees do NOT apply on Solana.)
MacroDroid relays Karoshi Discord posts here. Captured calls show up in the Degen → Alerts tab.
Managed by Claude via the MCP (its own single-TP + hard-SL, size set per trade, cap 1.0 SOL). No toggle here — Claude opens/manages it directly.
Auto-paper every @solearlytrending entry call (instant via the userbot push). 2% size · max 15 concurrent · retrace-limit entry (limit at call price −20%, valid 30 min — no market-chase; unfilled = skipped) · TP 1.4× full exit + hard SL 0.40× · holders ≥ 300 gate + rug gate (Security 🚨 / liq < $5K / fake ≥ 50% / on-chain veto). Forward-only. Exit model reworked v213.365 (eval 2026-07-06) — the tunable Karoshi exit ladder above does NOT apply to this lane.
Human-KOL #degens posts (relayed via MacroDroid) are parsed + LLM-classified; a token that builds momentum (repeat mentions / 🟢-buy escalation / Murad-list+buy / GMGN-flow corroboration) is flagged as a forming narrative. Alert-only — no auto-trade. Live narratives appear on the dashboard card. Note: these memecoins live for weeks/months, so any future auto-trade must NEVER force-close on a fixed hold-time.
MacroDroid macro: Discord #degens notification → POST this URL (include {not_text_big} for long posts). Distinct from the Degen Webhook above (that's the automated Karoshi-AI scanner; this is the human #degens channel).
WARN = detected narratives are only logged + shown on the dashboard card. LIVE = also sends a Telegram alert (category memecoin_narrative). Start in WARN, review the would-send log, then go LIVE.
Karoshi's #white-alert-24-7 Discord channel (relayed via MacroDroid). Two alert kinds are kept: WHALE positioning (BTC/alt leveraged perps → the dashboard Whale Positioning card + the agent's market context) and SMART MONEY (smart-wallet accumulation → degen narrative corroboration). General news is deliberately skipped (it duplicates the news feed). Read-only — never trades.
MacroDroid macro: Discord #white-alert-24-7 notification → POST this URL (include {not_text_big} for long posts).
Karoshi's Macro Intel Discord feed (relayed via MacroDroid): real-time macro / geopolitical headlines + economic prints. The bot ingests the headline + print result into the news pipeline (→ LLM agent context + dashboard), deduped against the other feeds (Al Jazeera / CNBC / Telegram) so overlaps are dropped — it fills coverage gaps like Reuters-sourced stories we otherwise miss. The per-asset impact arrows are stripped (opaque third-party prediction — not a signal). News only — never trades.
MacroDroid macro: Discord Macro-Intel notification → POST this URL (include {not_text_big} for the full post + footer).
A daily Claude-Code routine reads this bot's market intelligence over MCP and opens/closes paper swing trades in this scope. Demo only — the MCP write tools can never place a live order. View open trades under Swing in the navbar.
vectis with your bearer token:/mcp — the vectis server should list its tools: read (get_scoreboard, get_asset_detail, get_market_context, get_holdings) + demo-locked write (open_demo_trade, close_demo_trade).Authorization: Bearer <token> header into a custom HTTP connector in the Claude app.These settings apply to all demo accounts.
Restricts the New-Trade autocomplete (Demo mode), the /webhook/demo Signals-AUTO filter, and the Scanner Auto-Trade watchlist.
Demo IVB and Demo Scalp have their own per-symbol whitelists with sizing hints — by design their methodology only applies to a curated subset, so they override this list.
Since v201.55 the list is derived automatically from your TradingView Watchlist on every sync (default daily 17:00 UTC, manual button under Settings → System → TradingView Watchlist Sync).
To add or remove symbols, edit the TV watchlist and run a sync — this list will follow.
Simulates real trading costs. Applied identically to both demo accounts.
Hide illiquid assets from the Alerts list (Live mode only). The New-Trade ticker search always lists every Lighter market and tags each row with its liquidity tier — the call is yours.
Telegram alerts when scanner detects score changes or bias shifts.
💡 Scanner Alerts work when the browser is open. For 24/7 alerts without a browser, enable Auto-Trade below — it runs its own server-side scans.
Bias Change Alert
Automatic Top-10 summary via Telegram.
Scan each asset across multiple timeframes for confluence confirmation.
Timeframe Preset
A read-only AI layer that reads the bot's own trade history and logs and reports back in plain language. It never places, closes or modifies a trade — it only observes and advises. All numbers are computed by the bot in Python; the model only interprets them and must cite its source for every figure.
⚠ No API key configured on the server — the analyst is inactive. Set ANTHROPIC_API_KEY and restart to activate.
News Importance — Haiku Re-Rank (on-demand)
Curates the RSS "Top News" feed (the market-context news block) with Haiku on top of the always-on heuristic — picks + orders the most market-moving distinct stories with a one-line why. Runs headless on your Claude subscription (no metered cost; falls back to the Haiku API only on a rate-limit), 60-min cache. Off = heuristic ranking only, no LLM call.
A short morning summary of the last 24h across all modes (Live + every Demo mode): win-rate, P&L, what closed how, and any veto/filter or risk-pause worth knowing. Every brief is archived (History); it goes to Telegram (category Reports) only when Telegram below is On - default Off. Rhythm: daily ~06:05 UTC. Model: Sonnet, run headless on your Claude subscription (no metered cost; falls back to the API only if the subscription is rate-limited).
A health watchdog that scans the last 24h of the log for the known failure patterns from past incidents (unprotected trades, order-placement collisions, SL/TP churn storms, tracebacks, rate-limit bursts) and alerts only when one is actually emerging. Normal operation stays silent. The bot's real-time safety alerts are unaffected — this is an extra net for slow-building patterns. Telegram (category System). Rhythm: once daily ~00:10 UTC. Model: Sonnet, run headless on your Claude subscription (no metered cost; API fallback only on a rate-limit).
A deeper 7-day review with week-over-week shifts, per-mode structure, and up to 3 concrete tuning proposals (each with rationale, evidence and a confidence rating). Proposals are reported only — you decide, nothing is applied automatically. Every deep-dive is archived (History); it goes to Telegram (category Reports) only when Telegram below is On - default Off. Rhythm: Mondays after the daily brief. Model: Opus, run headless on your Claude subscription (no metered cost; API fallback only on a rate-limit).
Claude decides a direction (long / short / skip) for ONE symbol from the bot's own market context (scanner, proxy regime, Fear&Greed, funding) plus your note, sizes by confidence, and PROPOSES a trade in the Agent demo scope — you confirm before it opens. Ask Agent now lives in the New tab — set the top pill to Agent and open New. Master toggle here: Off = proposals only (Execute disabled); On = Execute enabled. Model: Opus — runs headless on your Claude subscription (no per-decision cost; falls back to the metered API only on a rate-limit or a malformed response). The agent now reasons from many more market-context sources — see below. Manual Ask-Agent is demo-only; live auto-trading is the separate Demo/Live toggle below.
Watches the autonomous routine arms on discarded/held setups never alert you (only watches you set yourself in the MCP chat do). When ON, a routine watch that fires re-invokes the routine to decide open (market/limit) / keep-waiting / discard on that setup — through the normal escalation + Opus red-team + risk gates (demo/paper, never live). When OFF (default), a routine watch fires silently (logged, no alert, no trade). Your own MCP-chat watches are unaffected — they still re-eval + notify you.
Breaking news in the realtime Telegram feeds (war escalation, energy-supply shocks, emergency central-bank action, systemic crypto events) triggers an immediate routine run instead of waiting for the next cadence slot. Guarded: keyword pre-filter → 5-min headline burst → one Haiku classification → fires only if a mapped asset actually moved (e.g. WTI ±0.8%, BTC ±1%) since the burst opened — fake or stale headlines never trigger. Max 3 fires/day, 90-min cooldown per event class. The run protects open positions first, then evaluates beneficiaries — with an explicit no-chase rule. Demo/paper only, normal escalation + red-team + risk gates. Default OFF.
When ON, the agent analyses incoming signals and scanner score-crossings on the whitelisted symbols and — on an open decision at or above the min confidence — opens an Agent trade automatically, no confirm. Skips during blackouts; a per-symbol cooldown and a daily cap bound how often it runs. Each decision runs headless on your Claude subscription (no metered cost; API fallback only on a rate-limit). Default OFF. Mode: Demo reacts to /webhook/demo (Signals) + scanner and trades the demo account; Live reacts to /webhook/alert + scanner and will place real orders — the live executor is a separate, reviewed build, so Live currently analyses but does not place trades yet.
Every ingested TradingView Daily Briefing is archived now. Browse the full history to track market context over time.
Daily check (free metadata call — no tokens) that the configured Claude models still exist, plus a hint when a newer one ships. Retirements or mid-call rejections also alert via Telegram (category System).
Estimated metered API spend from the call log (list prices; actual billing is on the Anthropic dashboard). Calls that ran headless on your Claude subscription cost nothing — counted separately as “free” next to each call count.
The bot triggers a fresh headless Claude on a schedule (day 06–20 UTC hourly / night every 3h) + on market events (swing fill, SL/TP close, lead-proxy regime-flip, scanner score-jump ±20, blackout). Each run passes through guards (single-flight · rate-limit · per-day cap · cautious-mode at -2% NAV halves size) → the demo-locked routine.
DEMO / PAPER ONLY — never touches live funds. Dry = analysis + alerts (no writes); Live = opens/manages demo-locked paper trades. Default OFF.
Weekend dead-zone Fri 20:00 → Sun 22:00 UTC (TradFi closed) runs at this sparse cadence; event-triggers still fire normally.
When swing-demo PnL is down ≥ this % of NAV on the day, the routine does NOT stop — it enters cautious mode for the rest of the UTC day: effective size is halved (×0.5) and it must review the day's losing setups before any new entry. The hard daily-loss halt stays the separate Risk setting.
Recommended rollout: enable Dry first, watch the routine journal + Telegram alerts for a day, then switch to Live and soak 48–72h before relying on it.
Checking engine…
The mic in the Ask-Agent chat. German by default. 'Accurate' uses a larger model with clearly better German, but it takes ~8-10s per phrase on this server; 'Fast' is ~2-3s.
Fast: base model — ~2-3s per phrase. Switch to Accurate if German misreads bother you.
Bot is running normally and monitoring all positions.
Hide the Alerts tab when you do not use it (it lost most of its purpose when the Signals mode was archived). Hiding it also frees a slot in the phone navigation bar, so Cycle moves out of the hamburger menu and into the bar. Nothing is deleted — alerts keep working, and switching this back restores the tab. Per device.
Show Stop Loss, Take Profits and the two funding rows in active trade cards. Applies bot-wide on every device; the per-card eye icon still overrides it per device.
Show or hide the small action buttons on the trade cards in the Trades tab. Default: all hidden (the close button always stays). Applies bot-wide on every device.
Sync the bot with your active positions on Lighter.
Pulls your public TV watchlist by share-ID and writes it to data/user_watchlist.txt.
Auto-runs daily at 17:00 UTC (1h before the Weekly Watchlist Review).
Telegram sends a diff on change and an error report if the sync fails.
Last sync: never
Restart the bot process. Useful if the bot seems stuck. Total downtime ~12-15 seconds (systemd auto-restart). Open positions on Lighter stay intact — SL/TP orders remain active during the restart.
Automatically install Lighter SDK updates when no live trades are active.
Pop-up toasts in the dashboard when an order is filled or a TP / SL is hit — across ALL modes incl. LIVE, each prefixed with the mode name (e.g. LIVE 🛑 BTC SL …). Independent of the Telegram notifications below; toasts only appear while the dashboard is open.
Choose which events trigger a Telegram message. Error alerts are always sent.
Error alerts are always sent regardless of the settings above.
All saved versions of the bot.
Are you sure you want to log out? You will need to re-enter your password and 2FA code to sign back in.
Scan this QR code with your authenticator app (Google Authenticator, Authy, etc.)
Or enter this code manually:
Emergency Backup Codes — save these now!
Each code can be used once instead of your authenticator. Store them safely offline.
Sessions stay active for 30 days — applies to web UI access only. SSH and TradingView webhooks are unaffected.
The bot process will exit and systemd will bring it back up automatically.
Downtime: ~12-15 seconds
Open positions on Lighter stay intact — SL/TP orders remain active during the restart.
Register the REAL contract for this ticker. Any other mint with the same ticker is then flagged as an impostor + demoted. Verify on GMGN first (use the card's copy button).